A few years ago, I designed a multiplexer that interfaced four T1 twisted-pair telephony lines to optical fiber. The integrated-CMOS T1 line drivers automatically generated the correct analog pulse waveshape to satisfy the Bellcore pulse mask requirements. Their only problem was they were somewhat picky about the termination impedance: More than 10% termination mismatch away from 100Ω and they complained by producing a distorted pulse shape that no longer agreed with the mask. The greater the impedance mismatch, the more distorted the pulse shape. This behavior was not a big deal; it is usually easy to match line and termination impedances to better than 10%.

All our prototypes were working fine until the day we obtained some T1 repeaters from a well-known manufacturer and attempted to drive their receivers with our prototypes. Our transmitted pulse waveform was almost unrecognizable on the scope; it looked just like it was driving into an open circuit.

I called the repeater manufacturer, and a support engineer assured me that the repeaters had an input impedance of 100Ω. He added that, when their other customers complained about similar problems, the standard approach was to tell them to add a 100Ω resistor across the receiver input.

I pointed out that this approach would result in a 50Ω termination and make things worse. If other customers were having similar problems, then maybe something was wrong with the repeater design. The engineer insisted that there was nothing wrong with it; all their lab tests had shown that the “active termination circuitry” worked properly. But he did say he would look into it some more.

In the meantime, my boss was somewhat upset that I could not make our product work with these repeaters. I demonstrated that our product worked nicely with 100Ω termination, and because it did not work with the repeaters there must be a repeater problem. The boss could not accept this hypothesis.

About a week later, the repeater manufacturer engineer called me back and explained that improperly loaded CMOS drivers could be responsible for the pulse distortion. He also explained that this problem had not shown up with older driver designs that used RC waveshaping networks. He then sheepishly admitted that the company’s production department had made a vendor change for a FET in the active input-termination circuit. The new FET package had a different lead arrangement that production didn’t notice at the time of substitution. Hence, all the repeaters the company had shipped since the vendor change did not present a 100Ω termination to the line, and production testing did not include input-impedance measurement. Oops!

I offered to fix our repeaters myself if he could send the new FET-pinout arrangement, but he insisted that we return the repeaters to the company for rework. We did, and our product then worked great.

The only problem we had after that situation was that a Baby Bell telephone company evaluated a product sample and complained about the transmitted pulse shape. Whoever had wired the central office made the mistake of using 72Ω twisted pair instead of the standard 100Ω cable.

You can reach design consultant Glen Chenier at glen@teetertottertreestuff.com. Like Glen, you can share your Tales from the Cube and receive $200. Contact edn.editor@reedbusiness.com.